Gas station without pumps

2020 May 17

Sixth video for electronics book

I’ve just published my sixth video for my Applied Analog Electronics book.  This video is on impedance spectroscopy, which is the measurement of impedance as a function of frequency.

I filmed the video using OBS (Open Broadcaster Software), and this is the unedited eighth take. I had a lot of trouble with the software or with losing the shorting wire. I have not figured out a really good way to use OBS with Waveforms, since the popup windows don’t get captured by OBS unless I use full-screen display.

2020 May 13

Fifth video for electronics book

Filed under: Circuits course — gasstationwithoutpumps @ 20:05
Tags: , , , ,

I’ve just published my fifth video for my Applied Analog Electronics book.  This video completes Chapter 28, giving the impedance for RLC circuits.

I filmed the video using OBS (Open Broadcaster Software), and this is the unedited first take.

2020 May 12

Fourth video for electronics book

Filed under: Circuits course — gasstationwithoutpumps @ 20:22
Tags: , , , ,

I’ve just published my fourth video for my Applied Analog Electronics book.  This video is for §28.4 of the textbook (Bode plots for inductor combined with capacitors).

I filmed the video using OBS (Open Broadcaster Software), and this is the unedited second take. The first take was far too long and I messed up about 18 minutes in. I’ll have to do a separate video for the second half of that attempt.

2020 May 3

Speed of signals in transmission lines

Filed under: Circuits course — gasstationwithoutpumps @ 20:40
Tags: , , ,

This post continues the previous one on transmission lines.  I ended that post with the input impedance for a transmission line loaded with Z on the output being \begin{bmatrix} \cos( \omega \sqrt{LC}) & j \sqrt{L/C} \sin(\omega \sqrt{LC}) \\ j \sqrt{C/L} \sin(\omega \sqrt{LC}) & \cos( \omega \sqrt{LC}) \end{bmatrix} \begin{bmatrix} Z\\1 \end{bmatrix}, where L and C are the lumped inductance and capacitance of the line and the vector is interpreted as the numerator and denominator of a fraction.

If we plug in \begin{bmatrix}1\\0\end{bmatrix} for a short-circuit load or \begin{bmatrix}0\\1\end{bmatrix} for an open-circuit load, we get input impedances of j \sqrt{L/C} \tan(\omega \sqrt{LC}) and 1/( j \sqrt{C/L} \tan(\omega \sqrt{LC}), respectively. The periodic pattern \omega \sqrt{LC}= n \pi corresponds to n roundtrips of a signal on the transmission line.

If we compute the inductance and capacitance of a transmission line from its geometry, we should be able to determine the speed of transmission in the line. For example, if the transmission line is a pair of parallel wires, each w long, then the period corresponds to a wavelength of \lambda=2 w and we can multiply the frequency by the wavelength to get the speed of propagation.

We can approximate the inductance of a pair of parallel wires of length w, radius r, and distance between centers d as L= \frac{\mu_0 w}{\pi}\ln(d/r), where \mu_0 is the magnetic constant 4 \pi 10^{-7} T m/A. Similarly we can approximate the capacitance of the pair of parallel wires as C= \frac{\pi \epsilon_0 \kappa w}{ \ln(d/r)}, where \kappa is the relative dielectric constant and \epsilon_0 is the permittivity of free space.

That gives us \sqrt{LC} = w \sqrt{\mu_0 \kappa \epsilon_0}, so that 1 = f (2w) \sqrt{\mu_0 \kappa \epsilon_0}, or velocity is f \lambda = 1/\sqrt{\mu_0 \kappa \epsilon_0}. But 1/\sqrt{\mu_0 \epsilon_0} is the speed of light in a vacuum c, so the speed of propagation f \lambda = c/\sqrt{\kappa}, which is the standard result for the propagation of signals in a transmission line.

Transmission lines

Filed under: Circuits course — gasstationwithoutpumps @ 13:20
Tags: , , ,

My book covers complex impedance of inductors and capacitors and looks a series and parallel circuits involving them—it even considers complicated non-linear models that have (j \omega \;1\,s)^\alpha M components for modeling loudspeakers (with \alpha\approx 0.6) and electrodes (with \alpha\approx -0.5). But I don’t consider models of distributed inductance and capacitance, commonly called transmission lines.  For the low-frequency work we do in bioelectronics, these more sophisticated models are not needed.

Model of a transmission line using 3 LC elements. The transmission-line model takes the limit of N elements as N goes to infinity, with each L_i= L/N and each C_i = C/N.

But I’ve been thinking about them sometimes at night, doing the algebra in my head as a way to get to sleep, so I thought I would make a record of my thinking.  It is fairly easy to determine the impedance of an infinite transmission line, by looking at what happens if you add a single inductor and capacitor at the front:

Treating the infinite transmission line as a single unknown impedance Z, we can solve an equation with Z as both the input impedance and the load impedance.

We can write the input impedance as Z = j\omega L + \frac{1}{1/Z_1 + j\omega C}. If Z_1 is an infinite transmission line, then Z_1=Z and we can simplify to the quadratic equation Z^2 - j\omega L_1 Z - L_1/C_1 = 0, which has solutions Z= (j \omega L \pm \sqrt{-\omega^2 L_1^2 + 4 L_1/C_1})/2. But we are taking the limit as N\rightarrow \infty with L_1=L/N and C_1=C/N, so this simplifies to just Z= \pm \sqrt{L/C}, which is the standard result for the impedance of transmission lines.

If we have a finite-length transmission line, but connect a load resistor to one end with resistance R=\sqrt{L/C}, matching the impedance of the transmission line, it looks at other end just like a resistor R=\sqrt{L/C}. But what happens if we attach a different load (say a short-circuit, an open-circuit, or an arbitrary impedance Z)?

I found it too difficult to do the algebra in my head for concatenating multiple LC units, until I remembered a trick I used in my very first paper (A proof of the isomorphism of wxyz-transformals and 2×2 integer matrices under multiplication. Computers and Mathematics with Applications, 7(5):425–430, 1981. scanned copy).  We can represent a fraction as a vector of the numerator and denominator, and multiply by 2×2 matrices to do simple transformations.  For example, if an impedance is represented as \begin{bmatrix} Z\\1 \end{bmatrix}, we can represent series connection as \begin{bmatrix} 1&Z_s\\0&1\end{bmatrix}  \begin{bmatrix} Z\\1 \end{bmatrix} and parallel connection as \begin{bmatrix} 1&0\\1/Z_p&1\end{bmatrix}  \begin{bmatrix} Z\\1 \end{bmatrix}.  If we look at putting one of our LC units in front of Z, we get \begin{bmatrix} 1-\omega^2 LC/N^2&j \omega L/N\\j \omega C/N&1\end{bmatrix}  \begin{bmatrix} Z\\1 \end{bmatrix}, and N of them is \begin{bmatrix} 1-\omega^2 LC/N^2&j \omega L/N\\j \omega C/N&1\end{bmatrix}^N  \begin{bmatrix} Z\\1 \end{bmatrix}.

That doesn’t look simpler, until you rearrange to get \left( I + \frac{1}{N} \begin{bmatrix} -\omega^2 LC/N&j \omega L\\j \omega C&0\end{bmatrix}\right)^N  \begin{bmatrix} Z\\1 \end{bmatrix} and then take limits as N\rightarrow \infty to get \left( I + \frac{1}{N} \begin{bmatrix} 0&j \omega L\\ j \omega C&0\end{bmatrix}\right)^N  \begin{bmatrix} Z\\1 \end{bmatrix}.

We can make the mental math easier if we make the matrix a root of unity:

\left( I + \frac{j \omega \sqrt{LC}}{N} \begin{bmatrix} 0&\sqrt{ L/C}\\ \sqrt{C/L}&0\end{bmatrix}\right)^N  \begin{bmatrix} Z\\1 \end{bmatrix}.  We can then use the well-known limit \lim_{N\rightarrow\infty} (1+x/N)^N = e^x, which works with matrices as well as it does with numbers.  We now have e^{\left(j \omega \sqrt{LC} \begin{bmatrix} 0&\sqrt{ L/C}\\ \sqrt{C/L}&0\end{bmatrix}\right)}\begin{bmatrix} Z\\1 \end{bmatrix}.

Because the square of our matrix is the identity matrix, the even powers accumulate on the diagonal and the odd powers accumulate in other two corners, so we can do exponentiation to get \begin{bmatrix} \cos( \omega \sqrt{LC}) & j \sqrt{L/C} \sin(\omega \sqrt{LC}) \\ j \sqrt{C/L} \sin(\omega \sqrt{LC}) & \cos( \omega \sqrt{LC}) \end{bmatrix} \begin{bmatrix} Z\\1 \end{bmatrix}.

We can do a simple sanity check by plugging in Z= \sqrt{L/C}, which yields \begin{bmatrix}\sqrt{L/C} (\cos(\omega \sqrt{LC})+ j\ \sin(\omega \sqrt{LC})\\ \cos(\omega \sqrt{LC})+ j\ \sin(\omega \sqrt{LC}\end{bmatrix}, which yields the input impedance \sqrt{L/C}, as desired for a transmission line with a matched load.

We can now plug in \begin{bmatrix}1\\0\end{bmatrix} for a short-circuit load Z=0 or \begin{bmatrix}0\\1\end{bmatrix} for an open-circuit load Z=\infty to get input impedances of j \sqrt{L/C} \tan(\omega \sqrt{LC}) and 1/( j \sqrt{C/L} \tan(\omega \sqrt{LC}), respectively.  At very low frequency (\omega \ll 1/\sqrt{LC}), where we would not normally use transmission-line models, we can use the approximation \tan(x) \approx x, which gives the low-frequency approximations j \omega L for the short-circuit load, and 1/{j \omega C} for the open-circuit load, which are just the approximations we would use if we were doing lumped models of a loop of wire or a pair of unconnected wires.

It is even more interesting to see frequencies where the input impedance goes to 0 or to infinity, which correspond to the periods for reflections off the far end of the transmission line.  We should be able to use the formulas for capacitance between two long parallel wires and inductance of a loop of two long parallel wires to get the L and C values, and then use the periods to determine how fast waves travel in the transmission line.  I’ve not done those calculations yet, as I have not memorized the formula for inductance or capacitance of parallel wires, nor the values of the permittivity of free space and the magnetic constant.  Ignoring the values of the constants, I think that dimensions all work, with both inductance and capacitance proportional to the wire length and with everything cancelling except the relative dielectric constant $\latex{\kappa}$, the permittivity of free space, and the magnetic constant.  I should work it out more carefully, though, to make sure I get the standard result that the velocity in a transmission line is c/\sqrt{\kappa}.

Writing up this description of transmission lines probably took me about as long as working out the matrix equations in my head did (a few hours).

Next Page »

%d bloggers like this: